01In plain English
A model can only consider a limited amount of text at a time, measured in tokens. Everything counts toward it: the prompt, any files you attach and earlier messages. When a conversation runs past the limit, older parts drop out.
02How it works
Think of the context window as the model's desk. Everything it can see for the current reply must fit on it: the instructions the tool gave it, your question, any documents you attached and the earlier messages in the conversation. The size of the desk is counted in tokens. When the desk is full, the tool has to take something off, usually the oldest messages, and the model no longer sees them, so it may forget a detail you mentioned an hour ago. A bigger window lets the model keep a whole contract, a long report or an entire conversation in view at once. A bigger desk is not free, though: longer inputs generally cost more and run slower, and models can miss details buried in the middle.
03Why it matters when you are choosing
A bigger context window means a tool can read a long contract or a whole codebase in one go. A small one forces it to work in pieces and forget earlier detail.
04What to check
Check the context size on your plan, since tools often give the largest window only on higher tiers.
05Where you will meet it
06Common questions
What happens when I go over the context window?
Older parts of the conversation or the extra text no longer reach the model, so it can forget earlier instructions or ignore the end of a long file. Some tools warn you, and some cut text silently. Start a fresh chat with a short summary when that happens.
Is a bigger context window always better?
Not always. A larger window lets you include more material, but it can cost more, respond more slowly and make it easier for details to be missed in the middle. For a short question, a small window is plenty, and the extra size would only add cost.
How many pages fit in a context window?
It depends on the model and the page. As a rough guide, a page of plain text is a few hundred tokens, so windows measured in tens of thousands of tokens hold dozens of pages, and larger ones hold whole books. Tables and code use more tokens per page.